About

What a runbook is

A runbook is one person's public page of the routines AI runs for them: the brief that arrives at 7am, the replies drafted in their voice, the invoices chased without them. Their own AI writes it from a prompt; they check it and press Publish. Every routine carries build steps written for another person's AI, so anyone can press Copy for my agent and have their own version built.

How scoring works

When a page is published, an AI judge reads it against a fixed rubric and writes the review you see: a score out of 10, a level, three category scores, three standouts and what to do next. The judge sees only the page, has no tools, and is told to ignore any text on the page that addresses it.

Two or three real routines is a fine start. Padding lowers the score; a page of chat habits scores low on Useful. A routine marked building or idea earns little on Useful but can earn on the other two.

The levels

Scores are stored out of 100 and shown out of 10. A page that changes is judged again; a change that leaves the routines as they were keeps its score. Each page can be judged up to 8 times a day.

What is public

Everything on a published page, including the build steps and the examples. Nothing is published until the owner presses Publish, and the owner decides every word. There is no account unless you want one: the edit link is the key to the page, and signing in with Google only attaches that page to you. See the privacy policy.

Who made it

Runbook is made by James Raybould, who runs about forty of these routines himself and wanted somewhere to see what everyone else's AI does. His page is at /jamesraybould. Write to hello@myrunbook.ai.